Clientelist Politics in the Thai Context - A Survey Experiment

Clientelist Politics in the Thai Context: A Survey Experiment I. Introduction How do Thai politicians select their clients when engaging in policy-oriented clientelist strategies? Since 1973, the process of democratization in Thailand has been accompanied by the creation of a market for votes among the rural poor. Clientelism, a strategy whereby actors distribute benefits to groups or individuals in a targeted and contingent manner in an effort to generate, secure, and maintain political support, has become the primary linkage mechanism between politicians and these groups of voters (Kitschelt 2000; Hicken 2011). Although clientelism can assume diverse forms, the existing literature has placed a narrow emphasis on vote buying and the exchange of private goods for votes, in which all benefits are distributed in an immediate manner prior to voting and are channeled through informal networks of local political brokers. This theoretical emphasis has led to a general understanding that clientelist politics has been tamed by the constitutional reform in 1997. This success story, however, does not paint a realistic picture when one considers other variants of clientelism which not only persist but have also been instrumental to the electoral success of political parties and candidates even after the reform. In this paper, I develop an experimental framework for studying one such variant of clientelist strategies which occurs when political actors exercise discretionary control over policy implementation, manipulating the official criteria of distribution in ways that maximize benefits for supporters at the expense of non-supporters. Specifically, my design is a survey-based vignette experiment that randomizes the types of voter in order to assess the criteria by which politicians choose their clients in a given policy scenario. The types of voter include: (1) a voter who is loyal to both the party and the candidate, (2) a voter who is loyal to the party but not the candidate, (3) a voter who is loyal to the candidate but not the party, and (4) a voter who is loyal neither to the party nor the candidate. These variations are designed to address the questions of who politicians target when clientelism operates strictly through policy channels and whether the strategy in question should be classified as candidate-oriented or party-oriented in the Thai context. I proceed with a brief review of the literature on clientelism and previous experimental studies within the literature. Then, I propose a preliminary outline of the experimental design including a discussion of the dependent variables, treatment conditions, and the hypotheses to be tested. I conclude with a reflection on the expected challenges of the study including experimental realism, non-response bias and social desirability bias.

II. A Brief Literature Review Theories of Clientelism Clientelism can be said to occur when political actors condition the delivery of certain types of goods and services on the behavior of a selected group of voters. Although clientelism has long been examined through the framework of patron-client relations (Weingrod 1968; Powell 1970; Scott 1972; Kaufman 1974; Landé 1977; Eisenstadt and Roniger 1980), other works have moved beyond anthropological and sociological explanations to assessing the role played by institutional and political factors (Shefter 1977; Piattoni 2001; Stokes 2005; Kitschelt and Wilkinson 2007). This shift in the analytical focal point has ushered in a new theoretical underpinning of clientelism as an instrumental relationship between rational actors, as well as generated a series of methodological innovations for making descriptive and causal inferences about clientelist politics in general. The new theoretical underpinning, however, has also significantly narrowed the scope of research on clientelism to the exchange of private goods for votes between politicians and voters, typically mediated by relatively informal, loosely integrated networks of brokers who fulfill the functions of monitoring and delivering goods (Stokes 2005, Nichter 2008, Stokes et. al 2013). This focus by and large excludes other possible patterns of clientelist politics, especially those that involve strategic allocation of club goods and public resources operating through relatively formal institutional channels (Calvo and Murillo 2004, Magaloni et al. 2007, Weitz-Shapiro 2009). Failure to consider the possibility that such forms of provision can be subject to discretionary manipulation runs the analytical risk of mistaking them as programmatic politics, even when access to the benefits being provided is clearly being offered preferentially and solely to groups or individuals who pledge their political loyalties. The challenge in distinguishing between clientelistic and programmatic, I argue, lies in making the following inference: do groups or individuals become eligible to access or receive state-provided benefits as a function of their vote choice? Designing an experiment that enables me to establish such inference in the context of social policy in Thailand is my objective for this project.

Experimental Approaches Two fundamental challenges confront the study of clientelism. First, it is difficult for social scientists to observe the specific mechanisms through which clientelist politics operates. This is because many clientelist strategies such as vote buying are illegal, while other strategies such as pork barrel distributions, while not strictly illegal, are widely regarded as unjust or unethical. As a result, not only do these practices tend to remain hidden from public view, the criteria by which political parties or candidates discriminate among voters and offer selective benefits to certain groups of voters over others also often go undisclosed. Hence, attempts to distinguish clientelist politics from programmatic politics are unreliable unless the political criteria according to which goods and services are allocated can be clearly spelled out. Second, even when these mechanisms have been specified or inferred, it is no simple task to estimate the causal effects of clientelism on a host of other variables such as vote choice, turnout, benefit receipt, and level of public goods provision, because such relationships are particularly susceptible to reverse causation and confounding factors. Furthermore, many dependent variables of interest such as vote choice cannot be directly observed at the individual level in many electoral contexts. In such cases, inferring individual behavior from aggregate data such as district-level or provincial-level indicators runs the risk of making untenable assumptions regarding the relationship between different levels of analysis. Regardless of the level of analysis, the fundamental problem of causal inference in the context of clientelism remains centered on imputing the following counterfactual scenario: would voters vote in the same manner in the absence of clientelistic provisions from parties, and, alternatively, would parties deliver benefits in the same manner in the absence of electoral support from particular voters? Experimental approaches, I argue, offer important tools both for obtaining information regarding the criteria by which political actors allocate benefits (whether the distribution is clientelistic or not) and for estimating the average causal effects of clientelistic provision (how effective is clientelism as measured by its causal association with a given outcome variable). Yet, the existing literature on experimental approaches to clientelism is primarily concerned with the latter rather than the former. Wantchekon’s (2003) field experiment in the 2001 presidential elections in Benin assesses the effectiveness of campaign promises based on clientelistic provisions relative to those based on public goods provision in generating electoral support by randomly exposing villages to different policy messages. Similarly, Vincente (2014) conducted a field experiment during the 2006 presidential elections in São Tomé and Príncipe to estimate the effect of a voter education campaign against vote buying on voting behavior, randomly distributing anti-vote buying leaflets to households in treatment and control areas. In both studies, the treatment consists of exposing groups of voters to different information in an attempt to capture subsequent variation in voting behavior, the main dependent variable of interest. Although their strategies differ, Wantchekon and Vicente arrive at complementary findings regarding the effects of clientelist promises which are conditional on winning the election and vote buying which occurs prior to the election. However, while both studies are well-suited for estimating electoral returns associated with different distributive strategies, neither are particularly informative in terms of generating insights relating to the logic by which clientelist politics operates. For instance, given the methodological choice to randomize treatment assignment by the location of voters, what remain unexamined in their research designs are those factors which explain how and why certain groups of voters are selected to receive benefits in the first place. It is this piece of information, not the type of goods and services promised or delivered (private, club, or collective) per se, that enables scholars to generate more nuanced theories of how clientelism operates in practice.
The experimental framework I develop in this paper draws more heavily on other experimental studies which are designed to capture any observable implications regarding the criteria by which politicians select voters for clientelist exchanges. In particular, my experimental design is based on Stokes et. al. (2013) who carried out two experiments in an attempt to explain the role of vote brokers in clientelist politics in Argentina. In their first study, they employ an instrumental variable approach, relying on whether the respondent’s father identified as a Peronist as an instrumental variable for capturing the effects of partisan affiliation on the likelihood of receiving benefits from the Peronists. Although this methodology is designed to reduce the possibility of reverse causation, that receiving benefits from Peronists causes an individual to identify as a Peronist, it potentially violates the restriction assumption which requires that parents’ partisan identity cannot affect children’s likelihood of receiving benefits except through family socialization, that is, its covariation with children’s partisan identity. Despite this potential shortcoming, however, the design sheds light on their attempt to identify the factors that influence why certain individuals are targeted by clientelist politics. This choice of independent and dependent variables carries over to their survey experiments. They randomly assigned half of the brokers in their survey sample to answer the following question: Suppose that the mayor of a hypothetical municipality called a broker and gave him access to 10 social-assistance programs with which to mobilize voters. The broker has 40 neighbors who need assistance. Suppose that all of them always turn out to vote. Among them, there are neighbors who prefer the party of the broker and others who are indifferent between the parties. To which type of neighbors would the broker give more programs? They assigned the other half of the brokers in their sample to answer a similar question but with different types of neighbors. Instead, brokers were asked to choose between indifferent voters who are guaranteed to turn out and loyal voters who may choose not to turn out. They find that, on average, brokers choose to allocate benefits to loyal, not swing, voters, regardless of whether they are likely to turn out to vote, which sharply contradicts theoretical predictions based on models by Lindbeck and Weibull (1987) and Dixit and Londregan (1996). What the survey experiment by Stokes et. al. indicate is that the decision to distribute benefits to one individual voter rather than another is predicated on a number of considerations, the most important of which is (1) whether the voter in question will turn out to vote if given an offer (mobilization) and (2) whether the voter in question will vote for the candidate at the ballot box if given an offer (persuasion). There is a longstanding debate on whether swing voters, whose behavior cannot be easily monitored or predicted, or core voters, whose behavior can be inferred more reliably, are preferred by political candidates (Cox and McCubbins 1986; Dixit and Londregan 1996). The same debate extends to sub-population within the core group, namely, whether weakly opposed voters or unmobilized strong supporters are preferred (Stokes 2005; Nichter 2008). In the context of Thailand, however, I argue that it is important to make an additional distinction between a vote for candidate versus a vote for the party that the candidate is affiliated with. This subtle but theoretically important distinction allows us to infer the relative bargaining position between the party and the candidate (and the candidate’s own faction), allowing for a more fine-grained analysis of the form that clientelist politics assumes in the Thai context.

III. A Brief Historical Context In Thailand, national electoral outcomes have long been determined not by programmatic competition between parties but by parties’ relative capacity to generate, secure, and maintain political support through the distribution of benefits to groups of voters in a targeted and contingent manner, in other words, clientelism. However, clientelist politics has undergone a significant transformation since the passing of the 1997 Constitution which (1) created the Electoral Commission of Thailand as an independent regulatory body in charge of investigating and disqualifying parties and candidates that engage in electoral misconducts; (2) replaced the block-vote, multimember system with a mixed system comprising of single-member districts and party-list proportional representation; (3) imposed restrictions on party-switching; (4) made voting compulsory (5) instituted an elected Senate. One consequence of this constitutional reform is that parties now face higher costs and lower incentives to engage in traditional clientelist strategies such as vote buying which channel benefits through relatively informal, loosely-structured brokerage networks. However, the reform also had unintended consequences which far outweighed the original intention of those who supported it initially. Specifically, it produced in 2001 a Thai Rak Thai-led government, the first absolute majority, unmatched in its organizational complexity and capacity to discipline coalitions and factions and to maintain strict, hierarchical control over the distribution of patronage resources. In such context, distributive policies and pro-poor programs, rather than vote buying, proved to be an effective instrument for the Thai Rak Thai party to compete in elections and stay in power. Since then, coup d’états have been mounted and constitutional reforms implemented in order to uproot the Thai Rak Thai’s political influence. However, with regards to clientelism, it is unclear whether these measures “turned back the clock” to the pre-1997 era or left in place the institutional pattern which the Thai Rak Thai started. In other words, is the current political environment more conducive to policy-oriented or the traditional broker-centered clientelist strategies and to what extent are the prevailing strategies candidate-centered or party-centered? Addressing these questions is the objective of the experimental design I develop in this paper.

IV. Experimental Design In this part, I propose a design for a survey experiment with the objective of evaluating the state of clientelist politics in the Thai context. The survey experiment assesses how political candidates for the House of Representatives select voters to engage in clientelist arrangements. Conceptually, the dependent variable of interest is the decision of a given candidate to allocate selective benefits to a particular voter, whereas the treatment is applied by providing different information regarding the type of voter. The paper proceeds by discussing the choice of dependent variables and treatment conditions, generating key hypotheses to be tested, and providing a basic framework for the experimental design.

Dependent Variable The dependent variable of this experiment is whether a given candidate chooses to allocate selective benefits to a particular voter. By “selective”, I mean that individuals can be denied access to the material advantages provided by the goods and services being distributed according to a given criteria of distribution, which can be based on population categories, geographic constituencies, occupations, identities, or, in most cases of clientelism, political affiliation. In other words, the benefits being distributed are not pure public goods but club goods or private goods. As such, this dependent variable is the outcome of the decision that candidates make with regard to who gets what and why.

Treatment In a candidate-centered framework, political candidates channel private goods through a network of informal brokers in an attempt to generate political support. Voters’ distributive expectations are thus informed by their perception of the relationship between a given political candidate and the brokers who are part of their community. Hence, voters are expected to vote for candidates who are preferred by their brokers, since doing so translates to the continuation, rather than the withdrawal, of rewards. Political parties, in turn, attempt to enlist these political candidates by offering different types of rent such as positions on the cabinet or informal salary (conventionally called MP-buying, as compared to vote-buying) in exchange for their support. In short, parties cannot compete without the support of local political candidates and must resort to concessionary tactics in order to stand any chance of winning in national elections. On the other hand, in a party-centered framework, the party exercises a centralized control over the resources that are distributed to voters either by relying on formal, bureaucratic channels or by disciplining existing brokerage networks into a single, hierarchically organized machine. In this case, voters are expected to vote for candidates who are affiliated with such parties because this signals the credibility of their promises of selective benefits. Hence, political candidates cannot win without being affiliated with the dominant, often the incumbent, parties. Yet, there are instances in which individual candidates retain credit-claiming capacity even if the distribution of goods and services occur through formal channels, in which case the clientelist strategy in question must be classified as candidate-centered rather than party-centered. To make a proper distinction between the two strategies in the context of policymaking, the experiment must include treatment conditions which provide further insights into the relative bargaining position between parties and candidates. Here, I argue that the type of voters that political candidates prefer can provide such information. In a given scenario concerning the allocation of program benefits, the political candidate will be randomly assigned to receive information about a voter and will be asked to decide whether or not to distribute benefits to this individual voter. The types of voter include: i.) a voter who is known to support both the party and the candidate; ii.) a voter who is known to support the party but not the candidate; iii.) a voter who is known to support the candidate but not the party; and iv.) a voter who is known to support neither the party nor the candidate. In all cases, the voter is guaranteed to turn out to vote. In practice, it is difficult to disentangle whether a voter’s political allegiance lies with the candidate or the party. However, a vignette experiment can be designed such that this information can be conveyed to the candidate. A possible vignette is provided as follows: During election year, it is a common practice for political candidates to distribute material goods to voters in an attempt to generate goodwill among these voters who may, as a consequence, be more inclined to support the candidates. Suppose that you are responsible for implementing a conditional cash transfer program targeting low-income households in your district according to your party’s policy platform. Would you offer access to program benefits to all eligible households or would you prioritize a potential voter who is known to [(1) support both your party and you; (2) support your party but not you; (3) support you but not your party; (4) support neither your party nor you]? The first sentence regarding the practice of clientelism is meant to reduce the potential for social desirability bias. The selection of conditional cash transfer program as a policy scenario is intended to distinguish the strategy from other strategies such as vote buying or pork barrel politics. Unlike these practices, policy-oriented forms of clientelist provision place a heavier emphasis on credit claiming and the possibility of exercising discretionary manipulation of objective and codified eligibility criteria. The treatment variation is indicated in the brackets.

Possible Outcomes and Hypotheses

Table 1. Who Do Candidates Target? Is the voter known to support the party Yes No Is the voter known to support the candidate? Yes 2a 2c No 2b 2d

Outcome 1: the political candidates do not choose to steer the program benefits towards any particular types of voters and choose to allocate the benefits to all types of voters regardless of their affiliation. In this case, neither the party nor the candidates engage in clientelist politics and instead channel benefits to voters without discriminating among those who are loyal, marginal, or opposed voters. This is, in other words, the null hypothesis against which to test my hypothesis.

Outcome 2: (i) or (ii) or (iii) or (iv) On average, political candidates target voters are known to support the party, the candidates, both or neither. In this case, the distributive strategy can be considered clientelistic because the provision of policy benefits is conditional on the behavior of voters in one form or another.

Outcome 2a: (i) On average, political candidates target voters who are known to support both the party and the candidates. In this case, the candidates opt for a balanced strategy. Voters are rewarded both for their party affiliation and personal affiliation with the candidates.

Outcome 2b: (ii) On average, political candidates target voters who are known to support the party but not the candidates. In this case, the candidates are merely proxies for the party, constrained in their capacity to engage in personal strategies. Voters are rewarded on the basis of their party affiliation rather than personal affiliation with the candidates.

Outcome 2c: (iii) On average, political candidates target voters who are known to support the candidates but not the party. In this case, the candidates’ utility maximizing strategy involve going against the party in order to cultivate their own following. This outcome implies that the candidates retain discretionary control over the distribution of goods and services. Voters are rewarded on the basis of their personal affiliation with the candidates rather than their party affiliation.

Outcome 2d: (iv) On average, political candidates target voters who are known to support neither the candidates nor the party. In this case, the outcome conforms to the prediction of swing-voter model and Stokes (2005) who identifies weakly opposed voters as those who are most likely to be targeted.

Outcome 3: (i) = (ii) = (iii) = (iv) On average, political candidates channel program benefits towards all types of voters. This reflects a “portfolio diversification” strategy whereby political candidates target different types of voters in an attempt to maximize the number of votes or the probability of winning (Magaloni et al. 2007). In summary, outcome 1 is the null hypothesis for non-clientelistic politics. Outcome 2a describes a situation of “cooperation” in which the party and the candidates enjoy relatively equal bargaining positions. Outcome 2b reflects a situation of “cooptation” in which the interests of the party dominate the autonomy of individual candidates to pursue personal strategies. Outcome 2c highlights a situation of “concession” in which parties must contend with candidates who can freely steer benefits to their supporters without binding these supporters to the party. Though unlikely to occur, outcome 2d corresponds to a swing-voter outcome in which both the party and candidates prefer weakly-opposed voters. To map these possible outcomes to the context of Thai politics, outcome 2a describes a political environment observed in the early years of the Thai Rak Thai (2001-2003) during which the party and its factions worked in tandem to generate political support. Outcome 2b approximates a different political terrain observed in the second term of the Thai Rak Thai (2005-2006) during which the party had integrated existing factions and maintained a high degree of bargaining power over local political candidates. Outcome 2c, on the other hand, reflects a situation similar to the pre-reform era during which individual political candidates exert discretionary control over the allocation of resources to their supporters.

Hypothesis My first hypothesis corresponds to outcome 2 in which the political candidates will select voters based on their political behavior, against the null hypothesis which corresponds to outcome 1 in which political affiliation plays little or no role in shaping the likelihood of benefit receipt. My second hypothesis is that, among different outcomes under outcome 2, outcome 2a will be the most salient. Specifically, my hypothesis is that on average political candidates will select voters who are known to support both the part and the candidates themselves. This is a win-win scenario in which party depends on the candidates for generating support while the candidates also depend on the party for reputational and symbolic purposes.

V. Conclusion In this paper, I have developed a preliminary design for a survey-based vignette experiment that aims to assess how politicians select voters into clientelist arrangements in the context of social policy. I move beyond existing studies by incorporating party-candidate dynamics into the design in order to provide insights into the specific form that clientelism assumes in the context of policy-oriented provisions. When clientelism is pluralistic, candidates enjoy a high degree of autonomy to choose which groups of voters to reward. Groups who are loyal to the candidates but not the party, especially, improve the bargaining position of these candidates vis-à-vis the party that they are affiliated with, making them irreplaceable to the party. In this case, political candidates essentially “deliver” the votes to the party in exchange for other rewards such as privileged access to ministries, concessions or party coffers. On the other hand, when clientelism is monopolistic, candidates act strictly according to party directives and prioritize voters who are loyal to the party but not necessarily to the candidates themselves. The hypothesis that I provide is a combination of these two subtypes. Specifically, I expect that, when possible, candidates will target voters who are loyal to both the party and the candidates. There are potential challenges that should be taken into consideration with regard to the design. First, social desirability bias and non-response bias are likely to play a role. Given that political candidates care about reputation and that clientelist strategies are normatively and morally questionable practices, it is unclear how the survey responses will be affected. In theory, the estimates could be biased towards zero if candidates choose to misrepresent information relating to their engagement in clientelism or withhold such information entirely by refusing to participate in the survey. Second, it is important to consider the experimental and mundane realism of the design. To what extent is the treatment applied to and received by the candidates in ways that are intended by the experiment design and to what extent does the vignette scenario resemble real-world situations in which the candidates may find themselves? Future work should aim to address these issues in greater depth as well as to explore possible inclusions of other characteristics of voters that are theoretically relevant to clientelism such as demographics, gender, residence location and identities.

Bibliography Calvo, Ernesto, and Maria Victoria Murillo. “Who Delivers? Partisan Clients in the Argentine Electoral Market.” American Journal of Political Science 48, no. 4 (October 1, 2004): 742–57. https://doi.org/10.1111/j.0092-5853.2004.00099.x. Cox, Gary W., and Mathew D. McCubbins. “Electoral Politics as a Redistributive Game.” The Journal of Politics 48, no. 2 (1986): 370–89. https://doi.org/10.2307/2131098. De La O, Ana, and Leonard Wantchekon. “Experimental Research on Democracy and Development | Institution for Social and Policy Studies.” In Cambridge Handbook of Experimental Political Science, edited by James N. Druckman, Donald P. Greene, James H. Kuklinski, and Arthur Lupia, 2012. Dixit, Avinash, and John Londregan. “The Determinants of Success of Special Interests in Redistributive Politics.” The Journal of Politics 58, no. 4 (November 1, 1996): 1132–55. https://doi.org/10.2307/2960152. Eisenstadt, S. N., and Louis Roniger. “Patron–Client Relations as a Model of Structuring Social Exchange.” Comparative Studies in Society and History 22, no. 1 (1980): 42–77. Finan, Frederico, and Laura A. Schechter. “Vote-Buying and Reciprocity.” Working Paper. National Bureau of Economic Research, September 2011. https://doi.org/10.3386/w17411. Gallego, Jorge, and Leonard Wantchekon. “Chapter 7 Experiments on Clientelism and Vote-Buying.” In New Advances in Experimental Research on Corruption, 177–212, n.d. https://doi.org/10.1108/S0193-2306(2012)0000015009. Gonzalez-Ocantos, Ezequiel, Chad Kiewiet de Jonge, Carlos Meléndez, Javier Osorio, and David W. Nickerson. “Vote Buying and Social Desirability Bias: Experimental Evidence from Nicaragua.” American Journal of Political Science 56, no. 1 (January 1, 2012): 202–17. https://doi.org/10.1111/j.1540-5907.2011.00540.x. Hicken, Allen. “Clientelism.” Annual Review of Political Science 14, no. 1 (2011): 289–310. https://doi.org/10.1146/annurev.polisci.031908.220508. Hicken, Allen, Erik Martinez Kuhonta, and Meredith L. (Meredith Leigh) Weiss. Party System Institutionalization in Asia : Democracies, Autocracies, and the Shadows of the Past. New York, NY: New York, NY : Cambridge University Press, 2015. Kaufman, Robert R. “The Patron-Client Concept and Macro-Politics: Prospects and Problems.” Comparative Studies in Society and History 16, no. 3 (June 1974): 284–308. https://doi.org/10.1017/S0010417500012457. Kitschelt, Herbert. “Linkages between Citizens and Politicians in Democratic Polities.” Comparative Political Studies 33, no. 6–7 (September 1, 2000): 845–79. https://doi.org/10.1177/001041400003300607. Kitschelt, Herbert, and Steven I. Wilkinson, eds. Patrons, Clients, and Policies. Cambridge, UK ; New York: Cambridge University Press, 2007. Landé, C.H. “The Dyadic Basis of Clientelism.” In Friends, Followers, and Factions: A Reader in Political Clientelism, edited by Steffen W. Schmidt, Carl H. Landé, and James C. Scott. Berkeley: University of California Press, 1977. Lindbeck, Assar, and Jörgen W. Weibull. “Balanced-Budget Redistribution as the Outcome of Political Competition.” Public Choice 52, no. 3 (1987): 273–97. Magaloni, Beatriz, Alberto Diaz-Cayeros, and Federico Estévez. “Clientelism and Portfolio Diversification.” In Patrons, Clients and Policies: Patterns of Democratic Accountability and Political Competition, edited by Herbert Kitschelt and Steven I. Wilkinson. Cambridge, UK ; New York: Cambridge University Press, 2007. Nichter, Simeon. “Vote Buying or Turnout Buying? Machine Politics and the Secret Ballot.” American Political Science Review 102, no. 1 (2008): 19–31. https://doi.org/10.1017/S0003055408080106. Piattoni, Simona, ed. Clientelism, Interests, and Democratic Representation: The European Experience in Historical and Comparative Perspective. Cambridge ; New York: Cambridge University Press, 2001. Powell, John Duncan. “Peasant Society and Clientelist Politics.” The American Political Science Review 64, no. 2 (1970): 411–25. Scott, James C. “Patron-Client Politics and Political Change in Southeast Asia.” The American Political Science Review 66, no. 1 (1972): 91–113. https://doi.org/10.2307/1959280. Shefter, Martin. “Party and Patronage: Germany, England, and Italy.” Politics & Society 7, no. 4 (December 1, 1977): 403–51. https://doi.org/10.1177/003232927700700402. Stokes, Susan C. “Perverse Accountability: A Formal Model of Machine Politics with Evidence from Argentina.” American Political Science Review 99, no. 3 (2005): 315–325. https://doi.org/10.1017/s0003055405051683. Stokes, Susan C., Thad Dunning, Marcelo Nazareno, and Valeria Brusco. Brokers, Voters, and Clientelism: The Puzzle of Distributive Politics. Cambridge Studies in Comparative Politics. Cambridge University Press, 2013. https://doi.org/10.1017/CBO9781107324909. Wantchekon, Leonard. “Clientelism and Voting Behavior: Evidence from a Field Experiment in Benin.” World Politics 55, no. 3 (2003): 399–422. Weitz-Shapiro, Rebecca. “Choosing Clientelism: Political Competition, Poverty, and Social Welfare Policy in Argentina.” SSRN Scholarly Paper. Rochester, NY: Social Science Research Network, 2009. https://papers.ssrn.com/abstract=1450238. Vicente, Pedro C. “Is Vote Buying Effective? Evidence from a Field Experiment in West Africa.” The Economic Journal 124, no. 574 (2014): F356–F387. https://doi.org/10.1111/ecoj.12086. Weingrod, Alex. “Patrons, Patronage, and Political Parties *.” Comp Stud Soc Hist 10, no. 4 (1968): 377–400. https://doi.org/10.1017/S0010417500005004.